Computer Vision
Introduction
Computer Vision is dealing with processing and automatically interpreting image-data through computer algorithms. The adhesion of the technology basis for Computer Vision with Computer Graphics made it a necessity to build-up corresponding know-how at VRVis. Current application areas are medical images and city models where methods for segmentation and reconstruction are playing important roles.
Photogrammetric Reconstruction
When you take a picture with a camera, the three-dimensional information - the depth of the scene - is lost. Photogrammetry is the discipline of recovering this three-dimensional information from one or more two-dimensional images. Depending on the required precision, these images can be taken by a wide range of devices, from cellphone cameras to professional, calibrated high-end cameras.
The recovered 3D information can be used in a variety of applications: for building virtual objects for online auctions, for architectural or archaeological reconstructions. You can even scan whole cities for urban planning, movie productions, or video games.
Additionally, computer vision techniques can be used to extract semantic data - meaning - from the 3D information. Instead of delivering only measurements of a city, these techniques will tell you how many windows and doors a house has, where the fire hydrants are, and how many trees are in a specific street.
Real Time Sensing
Since digital cameras are already cheap enough to be included in almost every kind of electronic gadget, they can be an inexpensive source for information. While mostly this data is gathered only for human eyes, delivering snapshots and video clips from one person to another, they can also be used for all kinds of measurements. In your car, cameras and similar sensors can be used to determine whether you are driving to close to the next car or the curb, in logistics they can scan codes and destinations of parcels, and in computer games they watch your movements to let you interact with the virtual world inside the game.
All of these applications require that digital image streams - video clips - are analyzed quickly and efficiently. The extracted information allows computers to support and protect humans in situations where quick decisions have to be made in a rapidly changing environment.
Medical Image Analysis
The amount and size of images acquired by medical doctors and biomedical scientists in daily clinical routine as well as for research purposes increased steadily in the last decades. These often multidimensional and high resolution images contain a vast amount of information that has to be extracted and analyzed. The investigation of methods for automation and acceleration of these extremely time consuming tasks is subject of disciplines like digital image processing, object segmentation, quantification and labeling, and computer assisted diagnosis.
Computer Vision helps to automatically assess within seconds damages of the heart based on MR images after myocardial infarction, to detect malign forms of breast cancer, and to mark and quantify biological structures in high resolution microscopic images.
